Anthropic seeks 50.1% voting control for seven co-founders before IPO

By Splitfeed EditorialLast updated Sep 26, 2026, 19:17 UTC

On September 24, 2026, The Information reported that Anthropic is seeking shareholder approval for a supervoting share structure that would give its seven co-founders collective voting control of 50.1% of the company ahead of a potential IPO. The arrangement is modeled on Palantir's dual-class voting system. Each co-founder currently holds roughly 2% of the company's shares, meaning the proposed structure would grant voting power that far exceeds their equity stakes. Shareholder approval is required before the structure can take effect.

The proposal raises a question that sits at the center of AI governance debates: who should control the companies building the most powerful AI systems, and on what basis. Dual-class voting structures are not unusual in the tech industry, Google and Meta among others have used them, but Anthropic's case carries extra weight because the company has positioned itself as a safety-first laboratory operating under a public benefit mandate. Concentrating majority voting control in seven individuals, despite their minority economic stakes, means those founders could resist pressure from large investors, including the major technology companies that have poured billions into Anthropic, when making decisions about research directions, deployment, and policy. One co-founder has separately stated there is a better-than-even chance AI will be capable of building itself by 2028, a timeline that sharpens the stakes around who holds decisive authority.

The pro-AI camp has produced no published reactions to this story yet. Accounts that tend toward AI enthusiasm would typically frame founder control as a feature rather than a bug, arguing that mission-driven leadership insulated from short-term investor pressure is exactly what responsible AI development requires. The anti-AI camp has been vocal and broadly alarmed. Several accounts connected the governance move to a wider pattern of power consolidation. The account emoscientist noted that Anthropic has sought to restrict open-weight models while accumulating institutional control over the pace of AI development, calling this a Bond Villain problem. The account freddies described Palantir itself as a trojan horse for authoritarian control, and nivolet listed the supervoting structure as one component of a broader total-spectrum control infrastructure alongside military AI integrations and federal data architecture. The middle ground, represented here by davidjohansen, offered a partial defense, noting that Anthropic has stated it refuses to work on AI-controlled autonomous weapons, a distinction that camp treats as meaningful even if critics do not.

The vote to approve or reject the supervoting structure is the immediate event to watch. If shareholders approve it, the shape of Anthropic's governance heading into a public offering will be set, and the next inflection point would be the IPO filing itself, which would disclose the full terms of any share classes and the specific powers attached to founder votes.